* Enable auto-merge for translation-batches
* Attempt to approve and set auto-merge
* Temporarily disable check for tests
* Test commit. Revert later.
* Not self approving anymore
* Improve failure message
* Fix auto-merge
* Remove test code
* Add ownership to @github-actions
* one fast link checker to rule them all
* -L alias
* refactor
* hacking away
* passing pages
* adding a workflow
* limit level to critical
* ability to pass a specific set of files
* testing specific files
* remove temporary debugging
* reorganize the printing
* formatting of output
* check images too
* prettyFullPath()
* minimal permission on workflow
* update comment
* new permission
* fix actions shas
* Delete obsolete files from Crowdin in workflow
* Remove empty files before `crowdin upload`
Those files would have been rejected by Crowdin.
Close https://github.com/github/crowdin-support/issues/117
* Change delete command
* Add crowdin translations
* Run script/i18n/homogenize-frontmatter.js
* Run script/i18n/fix-translation-errors.js
* Run script/i18n/lint-translation-files.js --check parsing
* Run script/i18n/lint-translation-files.js --check rendering
* run script/i18n/reset-files-with-broken-liquid-tags.js --language=pt
* run script/i18n/reset-known-broken-translation-files.js
* Check in pt CSV report
Co-authored-by: Marcelo Jacobus <marcelo.jacobus@gmail.com>
* Experiment with making the tarball smaller
Part of #1248
* try this
* stop debugging
* delete translations too
* delete heavy search indexes too
* push and popd
* try this hack
* delete but leave directory
* debug more
* faster delete of translations
* less loud
* async await
* async await
* no tree
* simplify
* experimenting more
* unfinished
* only the large files
* change order
* brotli with level 6
* cope better with decorated rest json files
* tidying
* keep images
* cleaning
* cleaning up
* refactored function
* try this
* better comment
* remove console logging
* more important changes
* improved fs.accessSync check
The new order will use "reset known broken files" as a last step. That
way we can better understand if the previous steps were enough to catch
those errors. The least files are reset in that last step, the better.
Also, the check for broken/mismatched liquid tags is performed after
rendering and parsing check. That helps us to identify how useful the
mismatch check is, if at all.
* Add script to report reset files
* Has the workflow report the reset files
* Fix script name
* Filter report by language
* Update .github/workflows/create-translation-batch-pr.yml
* Use build in array methods
* migrate from actions/github-script to own scripts (staging deploy)
* make it executable
* remove need to --no-save install 'esm'
* don't use self-hosted runner so we get latest OS (2.285.0)
* remove pointless try/catch
